Beyond PLM examines eight competitive defenses for PLM software in the AI era

AI-generated summary

AI could make PLM applications easier to build and replace without eliminating manufacturers’ need for reliable product knowledge, change control, traceability and supplier coordination, argues Beyond PLM author Oleg Shilovitsky. His central thesis is that PLM’s operational importance may grow while its established business models face pressure. Drawing on two MicroConf talks by Rob Walling, he identifies eight potential competitive defenses based on trusted responsibilities rather than reproducible features. Shilovitsky discloses that he is OpenBOM’s co-founder and CEO; the article presents his analysis, not an announced product roadmap. AI-built alternatives illustrate both the threat and its current limits. PhotoCraft describes itself as early alpha, not a replacement for daily professional Photoshop work, while CADCraft’s pre-alpha roadmap acknowledges workflow and file-preservation gaps. Nevertheless, a challenger needs only enough functionality to solve a particular problem. Competition could come from customers, open-source developers, implementation partners, adjacent CAD or ERP platforms, and agents spanning multiple applications. Manufacturers might retain PLM records while needing fewer people to search, reconcile BOMs and coordinate handoffs. Free distribution is not new: Dassault Systèmes’ DraftSight attracted Bobst through its free application and paid Premium Service Pack, while Aras Innovator combined free software availability with subscriptions for support and upgrades. These examples show the value of dependable operations, but AI could also reduce implementation and integration effort, pressuring service revenues. The first two defenses are **industry trust and authoritative Product Memory**. Recommendations, responsive support, practical education and successful implementations create credibility that a new application cannot immediately reproduce. Product Memory extends formal records—structures, revisions, configurations and approvals—with decision context, rejected alternatives, accepted exceptions and subsequent outcomes. Replacing such a system requires preserving relationships and meaning, not merely importing files. Shilovitsky explicitly argues that this memory belongs to customers: defensibility should come from maintaining its integrity, accessibility and usefulness, not restricting portability. The third and fourth defenses concern **delegated action and dependable calculations**. An authorized agent could prepare an engineering-change package, check it, obtain approval, distribute the correct revision, collect acknowledgments and escalate exceptions. Its advantage would depend on reliable permissions, timing and accountability, with engineering decisions remaining assigned to responsible people. Similarly, trusted BOM cost and quantity calculations require correct configurations, consistent units, valid supplier prices and explicit assumptions—not simply arithmetic. Outputs should remain traceable and explainable even when users cease independently recalculating every result. The fifth and sixth defenses are **compliance assurance and active collaboration networks**. Regulated uses may require controlled records, approvals, traceability, testing evidence and validation support. The article cites the FDA’s February 2026 software-assurance guidance as an example of risk-based confidence in automation, while cautioning that serving regulated customers does not establish blanket approval or suitability. Collaboration across OEMs, suppliers, contractors and procurement can also make replacement difficult because participants must change shared practices together. However, dormant portal accounts offer little protection: a genuine network effect requires additional relevant participants to improve existing participants’ experience. The seventh defense is **learning from engineering and manufacturing outcomes**. In a hypothetical machined-housing change, the platform records the drawing revision, supplier clarification requests, approved interpretations, acknowledgments and production feedback. Recurring omissions, such as ambiguous surface-finish requirements, could inform earlier checks on future packages. Acknowledgment alone does not prove technical correctness; production results provide stronger evidence. Such learning must respect confidentiality and agreed rights, and qualification or field-performance history takes time to accumulate. The eighth defense, **maintained BOM and service information**, concerns ongoing reconciliation, updates, permissions and integration upkeep. Unlike an authoritative record’s role in establishing what is official, this service keeps information usable; it need not consolidate every source into one database. Commercially, Shilovitsky sees faster development, charging for agent use and delivering completed work as compatible approaches. Supplier-package preparation and approved distribution could become bounded services rather than tools requiring multiple licensed operators. Subscriptions, usage allowances, transaction fees or hybrids could support them, provided acceptance criteria and exception-handling costs are clear. He concludes that incumbents possess valuable records, relationships and experience, while entrants can reorganize these responsibilities. Neither has guaranteed protection: defensibility depends on what customers continue entrusting to the platform when alternative applications become easier to build.